On 6/9/2025 10:56 PM, Greg Mirsky wrote:
Hi Med,
thank you for leading this discussion. Please find my notes below
tagged GIM>>.
Regards,
Greg
On Fri, Jun 6, 2025 at 2:56 AM <[email protected]> wrote:
Hi all,
(changing the subject to ease tracking the discussion)
At least from where I sit, I think that that the inputs from
Andy/Matthew kindly helped to clarify the assumptions on the QoS
treatment.
Now, focusing on this part of the discussion where Greg suggests:
* In-flow OAM is an active or hybrid OAM method, as defined in
[RFC7799], that traverses the same set of links and interfaces and
receives the same Quality of Service treatment as the monitored
object.
Which is inspired by RFC9772:
* In-band OAM is an active or hybrid OAM method, as defined in
[RFC7799], that traverses the same set of links and interfaces and
receives the same Quality of Service treatment as the monitored
object. In this context, the monitored object refers to either
the entire Geneve tunnel or a specific tenant flow within a
given
Geneve tunnel.
I have some clarification questions for Greg:
* is this a proposal for replacement to the following terms or
are these new terms?
CURRENT:
Path-Congruent OAM:
The OAM information follows the exact same path as the
observed
data traffic. This was sometimes referred to as "in-band".
Non-Path-Congruent OAM:
The OAM information does not follow the exact same path
as the
observed data traffic. This can also be called Path-
Incongruent OAM, and was sometimes referred to as "out-of-
band".
GIM>> Yes, I propose "in-flow/out-of-flow" to replace
"Path-Congruent/Non-Path-Congruent" and
"Equal-Forwarding-Treatment/Different-Forwarding-Treatment" terms:
Equal-Forwarding-Treatment OAM:
The OAM packets receive the same forwarding (e.g., QoS)
treatment as user data packets. This was sometimes referred to
as "in-band".
Different-Forwarding-Treatment OAM:
The OAM packets receive different forwarding (e.g., QoS)
treatment as user data packets. This was sometimes referred to
as "out-of-band".
Those definitions centered around flow (in-flow/out-of-flow) are
actually not helping IMO
What is flow? I can provide some IPFIX references, but let's imagine the
typical 4 tuples to monitor the "user data packets", as mentioned in the
draft.
Now let's assume my IPFIX Metering Process has more Key fields for the
flow definition: QoS or the TCP flags are key fields.
Suddenly, I don't have one flow record, but multiple ones. Can we
guarantee that all these flow records will have the same treatment. Not
really.
Since you don't know the flow definition and you don't know the
forwarding device hashing function, simply telling: this should be same
flow, so should have the treatment is a simplification.
Regards, Benoit
_______________________________________________
OPSAWG mailing list -- [email protected]
To unsubscribe send an email to [email protected]
